Solved
Stratified sampling based on value of 10-item factor
Hi everyone,
For my master thesis I am using Qualtrics for a survey. I want to split my respondents into three groups that all have a percentage of high self-esteem people that is similar to the one general population (e.g. 15%).
The most common set of items to measure self-esteem has 10 items. Now, how can I score the results from these 10 items into one factor and use this value to assign the respondents to the three groups such that in the end, all three groups have similar distributions of high and low self-esteem people?
I have been unsuccessfully trying to find this out for a while now. Maybe someone can help? That would be fantastic!
Thank you and best wishes
Fred
Best answer by fleb
Hi @fhens,
as soon as your responses are submitted, you can access them using piped text from JavaScript. You can convert these values to numbers and do any math operations with them and determine if the value is high or low. Then you could branch your survey according to this embedded field and assign each branch separately to your groups.
However, this version doesn't take into account that high self-esteem respondents could leave survey with a higher probability if they're in one particular group. I don't know how likely is this since I don't know your survey. On the other hand, if you realize, that this happens, you can change the likelihood to be assigned to particular groups using "edit counts" (see manual for randomizers).
Nevertheless, if you'd like to use my suggestions, you could use something like this:
1) Javascript to get whether the is low or high
Add the following JavaScript to some question which is on some pages after the last self-esteem is answered, but before the first branching (see the Surve flow below). (If you need to do it on the same page where the last question was answered, it is also possible, but much more complicated.)
Qualtrics.SurveyEngine.addOnload(function()
{
var E = Number("${q://QID14/SelectedChoicesRecode}") + Number("${q://QID15/SelectedChoicesRecode}");
var esteem = (E >= 5) ? "high":"low";
Qualtrics.SurveyEngine.setEmbeddedData( "esteem", esteem );
});
The first line: an example of how to compute something based on questions answers; modify it such that you compute self-esteem.
The second line: set whether self-esteem is high or low using a given threshold
The third line: send the "low"/"high" value into the embedded field to future use in branching.
Note: if your questions are of a different type than multiple choice, the form of the corresponding piped text will be different. If you aren't sure what the form is, click on "Piped text" while editing some question, select the one you need and copy it to the proper place of the JavaScript.
2) Survey flow with proper randomizers and branching
!
!
!
!
Note: Block1 should not be there twice, block names are random here, just as an example.
!
!
!
Note: Block1 should not be there twice, block names are random here, just as an example.
+6Hi @fhens,
as soon as your responses are submitted, you can access them using piped text from JavaScript. You can convert these values to numbers and do any math operations with them and determine if the value is high or low. Then you could branch your survey according to this embedded field and assign each branch separately to your groups.
However, this version doesn't take into account that high self-esteem respondents could leave survey with a higher probability if they're in one particular group. I don't know how likely is this since I don't know your survey. On the other hand, if you realize, that this happens, you can change the likelihood to be assigned to particular groups using "edit counts" (see manual for randomizers).
Nevertheless, if you'd like to use my suggestions, you could use something like this:
1) Javascript to get whether the is low or high
Add the following JavaScript to some question which is on some pages after the last self-esteem is answered, but before the first branching (see the Surve flow below). (If you need to do it on the same page where the last question was answered, it is also possible, but much more complicated.)
Qualtrics.SurveyEngine.addOnload(function()
{
var E = Number("${q://QID14/SelectedChoicesRecode}") + Number("${q://QID15/SelectedChoicesRecode}");
var esteem = (E >= 5) ? "high":"low";
Qualtrics.SurveyEngine.setEmbeddedData( "esteem", esteem );
});
The first line: an example of how to compute something based on questions answers; modify it such that you compute self-esteem.
The second line: set whether self-esteem is high or low using a given threshold
The third line: send the "low"/"high" value into the embedded field to future use in branching.
Note: if your questions are of a different type than multiple choice, the form of the corresponding piped text will be different. If you aren't sure what the form is, click on "Piped text" while editing some question, select the one you need and copy it to the proper place of the JavaScript.
2) Survey flow with proper randomizers and branching
!
!
!
!
Note: Block1 should not be there twice, block names are random here, just as an example.
!
!
!
Note: Block1 should not be there twice, block names are random here, just as an example.Wow, thank you for your quick and detailed response, @fleb!
Actually I haven't been able to figure that out yet how to combine the items into the esteem embedded field. So if you could tell me how I would be super grateful!
Also, did you
- forget a randomizer after the low esteem if statement?
- in purpose leave the randomizer at "randomly present 3 of the following elements"?
- in purpose not tick the box "evenly present elements"?
I feel like all three of them were not in purpose because you just wanted to show the branching logic. But since I'm new to Qualtrics and since software can be super weird sometimes, I thought it'd be better to ask 😃
+6Hi @fhens,
you're right in all your three points. I have been finishing my response in a hurry, but wanted to post it to prevent someone else to unnecessarily do exactly the same work. I'll correct it also in my original answer to be more clear for eventual future readers and include there also the part about how to get "low"/"high".
Actually, if you randomly present any number of elements including 3, it will work anyway since the value of the field will be the last one. However, of course, it is unnecessary to select more than one element.
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.